Skip to main content
Queen Mary Academy
  1. Queen Mary Academy
  2. Educators
  3. Innovation and Scholarship
  4. Innovative Pedagogies
  5. Centre for Excellence in AI in Education
  6. Blog
  7. Items

Reframing Feedback: Using AI to Support the Affective Dimension of Learning

Dr Manesha Peiris outlines how feedback is central to learning but often carries significant emotional weight, particularly in high-pressure STEM contexts where grades are closely tied to learners’ sense of performance and validation. Feedback must be actionable, psychologically safe, and emotionally attuned, with LLMs offering potential support for more equitable learner engagement.

 

 

 

Published:
Profile photo of Dr Manesha Peiris

The Emotional Weight of Feedback 

While feedback is essential to learning, it often carries significant emotional weight. For many learners, assessment feedback is not neutral, but an affective experience, where comments intended to support improvement may trigger anxiety, defensiveness, or disengagement. This is particularly evident in high-intensity environments like STEM disciplines, where performance is closely tied to grading metrics, and learners may rely on grades as external validation. When feedback challenges expectations, it can create dissonance between self-perception and evaluated performance, disrupting engagement with learning. 

Research suggests learners do not always act on feedback due to how it is experienced (Lizzio and Wilson, 2008). Emotional responses can interrupt the cognitive processes needed to interpret and apply feedback. If feedback is inherently emotional, it also represents a site for affective development. This requires feedback to be interpretable, actionable, and psychologically safe, though structural pressures (Henderson et al., 2019) make this difficult. Within this context, LLMs offer potential as supportive tools for more equitable and emotionally attuned engagement. 

The Role of LLMs: Translating Feedback into Usable Insights 

LLMs like ChatGPT and Copilot can act as intermediaries between feedback and learner understanding. Their role is not to replace academic judgement, but to mediate how feedback is received, interpreted, and acted upon. This can empower learners to engage with feedback more constructively, particularly when it is experienced as critical or emotionally challenging. 

This article considers two complementary uses of LLMs: supporting educators in translating feedback into forms that are equitable, student-centred, and emotionally attuned, and enabling learners to reinterpret feedback through structured prompts. In doing so, LLMs can help make feedback more accessible, constructive, and actionable, supporting more effective engagement. 

For example:

Prompt: Rewrite the following feedback in supportive, constructive, and student-centred language. Keep the academic meaning intact, but make the wording clear, specific and emotionally safe. Focus on how the student can improve rather than what is lacking.  Feedback: “This argument lacks clarity”. 
 
Response: Here’s a supportive, student-centred rewrite that keeps the meaning intact while giving the student a clear way forward: 
 
Rewritten Feedback: 

“Your argument will come across more strongly if you explain your point in a clearer, more detailed way. Taking a moment to break down your reasoning step by step will help the reader follow your thinking and understand the conclusion you’re drawing.” 

Structuring Actionable Next Steps 

A common limitation in feedback practices is that markers may identify issues or gaps in performance without clearly articulating how these can be addressed. LLMs offer potential here by scaffolding feedback into structured, actionable guidance, enabling learners to effectively translate critique into improvement. For instance, LLM can: 

  • Breaking feedback into manageable steps 
  • Prioritising areas for revision 
  • Providing examples of improved work 

For instance, consider the academic feedback below:

Prompt: Rewrite the following feedback in supportive, constructive and student-centred language. Keep the academic meaning intact, but make the wording clear, specific and emotionally safe. Focus on how the student can improve, rather than on what is lacking.  

Feedback: “At present, references to established project estimation models are limited and described in vague terms, which leads to assumptions rather than clearly supported arguments 

Context: This is for a higher education project management module where students are expected to discuss cost estimation models such as COCOMO and Function Point Analysis. 

 Response: Here’s a supportive, student-centred rewrite that keeps the academic meaning intact while guiding the student towards improvement. 

Rewritten feedback: “To strengthen your argument, you could expand your discussion of established estimation models such as COCOMO or Function Point Analysis. Offering clearer explanations of how these models work – and linking them directly to your project context – will help you move from assumptions to well-supported, evidence-based reasoning.”

Examples 3 and 4 text: 

Prompt: Explain this feedback in simple terms, rewrite it in a supportive way, and give me clear steps and priorities to improve. At present, references to established estimation models are limited and described in vague terms, which leads to assumptions rather than clearly supported arguments.  

Response: Here’s a clear, student-friendly breakdown that explains the feedback, rewrites it supportively, and gives you a practical improvement plan you can hand straight to a learner. 

What the feedback means in simple terms 

The marker is saying: 

  • You mentioned project estimation models, but only briefly 
  • The way you described them was too general, so it’s hard to see exactly what you mean. 
  • Because of this, your argument feels based on assumptions rather than solid evidence.  

 Supportive, student-centred rewrite 

“You can strengthen your argument by expanding your discussion of established project estimation models and explaining them in clearer, more specific terms. Showing how these models apply to your project will help you build a well-supported, evidence-based argument rather than relying on assumptions.” 

 Clear Steps to Improve (with priorities) 

 Priority 1 – add specific models 

  • Choose one or two specific estimation models (e.g. COCOMO, Function Point Analysis) 
  • Name them clearly and explain why they are relevant to your project. 

 Priority 2 – Explain the models in concrete terms 

 For each model you use: 

  • Briefly describe how it works 
  • Identify what input it requires 
  • State what kind of estimate it produces 

 This shifts the student experience from: 

I don’t know what to do with this → I know what to try next. 

Supporting Reflective Engagement 

LLMs can support affective development by encouraging reflective engagement with feedback. For example, they can generate prompts such as:  

  • What do I understand? 
  • What is unclear?  
  • What should I change?  

These encourage learners to process emotions, develop self-awareness, and build agency in responding to feedback. 

A key concern is that AI may “soften” feedback. However, the aim is not to remove the challenge, but to reduce unnecessary threat. LLMs can help maintain academic rigour by framing critique as an opportunity, separating the learner from the work, and supporting growth-oriented interpretation. 

Used thoughtfully, LLMs act as scaffolds rather than substitutes. They can support feedback translation, embed reflection, and model constructive language, while preserving the educator’s original feedback. 

Towards Emotionally Intelligent Feedback Practices 

As we continue to explore the role of AI, its potential extends beyond efficiency and automation. One of its most significant contributions may lie in addressing a longstanding challenge: supporting learners to engage meaningfully with feedback, while acknowledging and reducing the affective barriers, such as anxiety and stress, that can hinder this engagement. By recognising feedback as both a cognitive and emotional process, and by using LLMs to support interpretation and reflection, educators can begin to design feedback experiences that are not only informative but also transformative. In doing so, feedback becomes more than a post-assessment ritual; it becomes a structured opportunity for affective and academic development. 

While the use of LLMs in assessment feedback presents significant opportunities, it is not without risk. One key concern is the over-sanitisation of feedback, which may diminish academic precision. If not carefully prompted, LLMs may prioritise tone over substance, potentially diluting important critique provided by the marker. There is also a risk of increased learner dependency on AI to interpret feedback. Over-reliance on such tools may erode learners’ ability to independently interpret and respond to feedback, not only in academic contexts but also in broader professional and personal settings. This, in turn, may limit the development of resilience required to engage with challenging or critical feedback. Furthermore, unless carefully monitored, LLMs may introduce misinterpretation or distortion of feedback. As these systems generate responses based on patterns rather than true understanding, there is potential for subtle shifts in meaning that may not accurately reflect the marker’s intent. Finally, while AI has the potential to improve efficiency in marking and feedback processes, it may also introduce a workload paradox. The use of LLMs requires educators to review and refine AI-generated outputs, which can add an additional layer of effort. This may also create pressure to produce increasingly “perfect” feedback, further complicating the workload rather than reducing it. 

Author Note 

This piece forms part of ongoing research into the role of feedback in supporting affective development in higher education. 

Bibliography: 

Dr Manesha Peiris

Lecturer in Reflective Practice and Project Managament, Director of Student Experience

https://www.qmul.ac.uk/spcs/staff/academics/profiles/mpeiris.html 

 

 

Back to top